Chapter 5 126 noise levels were measured in Chinese trains, varying from 88 to 93 dBA (Peng et al., 2019). The significant effect of train type and driving speed on the noise level confirmed the results of Kurze et al. (2000) and (Pronello, 2003). According to ISO 7731, warning signals will be clearly detectable for normallyhearing employees if the signal energy in one or more 1/3 octave bands is more than 13 dB higher than the effective-masked-threshold (ISO, 2003). The Detectsound analysis showed that this criterion is fulfilled for the DSD signal in all trains, except for the Mat64 when driving at maximum speed. For the ATP, this criterion is not fulfilled in most trains when driving at maximum speed. The ISOmethod only takes into account the highest signal component and ignores other spectral components. Using the Detectsound criterion of having at least three signal peaks within the design window for detection, we anticipated on the fact that detection of tonal signals may be better when multiple spectral peaks are more than 13 dB higher than the effective-masked- threshold (Edworthy, 1994). The lack of spectral elements within the design window does not necessarily mean that the signal is inaudible, which is shown by the results of the laboratory measurements. The SNR50 varied from -47.4 to -23.4 dB SNR, indicating that the warning signals remain audible when adjusted at levels significantly below the level of the background noise. Contrary to the Detectsound analysis, the outcomes of the laboratory measurements were more favourable for the ATP signal than for the DSD signal in most trains. A reason might be that we averaged the ATP measurements over the first 200 ms, although the sound level of the ATP signal decays over time. The higher onset of the ATP signal might result in better signal detectability than the analysis of the short-time averaged level suggests. Another explanation is that Detectsound may underestimate signal detectability when signal peaks are present at higher frequencies. According to ISO 7731, it is advised to include warning signals with signal components in the range of 500 to 2500 Hz when designing a warning signal (ISO, 2003). In line with this recommendation, the Detectsound model does not take spectral elements above 3100 Hz into account, anticipating on employees with high-frequency hearing loss due to presbycusis and/or noise exposure (Giguère et al., 2008). For the DSD signal, the prominent signal peaks were present between 500 and 4000 Hz and thus fall mostly within the spectrum of the design window. Contrary, the most prominent peaks of the ATP signal were present between 4000 and 8000 Hz.
RkJQdWJsaXNoZXIy MTk4NDMw